The current state of AI in healthcare and the problems it presents, the impact on clinical decision-making in hospitals and doctors’ offices, and why it is important for enterprises to put in place an AI governance system.
Introduction: What is Artificial Intelligence?
Research into Artificial Intelligence (AI) has been ongoing for decades, with early proposals dating back to 1950. However, only in recent years, it has seen a resurgence in popularity thanks to the increased availability of computing power and the growth of big data and machine learning. AI is the ability of machines to perform tasks that ordinarily require human intelligence, such as understanding natural language and recognizing objects. With the rapid expansion of AI, there are opportunities for businesses and individuals alike to capitalize on its capabilities.
AI is a field of computer science and engineering focused on the creation of intelligent agents, which are systems that can reason, learn, and act autonomously. It is defined as the ability of machines to perform tasks that ordinarily require human intelligence, such as understanding natural language and recognizing objects. In layman’s terms, AI can be used to create intelligent machines that have the ability to learn and work on their own.
With this technology becoming more and more prevalent in our lives, an increasing number of companies are starting to develop applications and products based on artificial intelligence. To date, there are already several commercially successful products that use artificial intelligence, however it is still an immature market and therefore there are opportunities for both companies and individuals to leverage the capabilities of AI to do business.
You might also like to read Model Operations for Secure and Reliable AI
Why AI is the Future of Medicine
Artificial intelligence is often spoken about in the context of technology and its future role in our lives. However, what many people don’t realize is that AI has the potential to revolutionize practically every industry, including healthcare. From diagnosing diseases to helping with surgeries, AI can make enormous contributions to improving patient care. Doctors are already using AI to improve their practices, and the future looks very bright for this technology. With continued advancement and implementation, AI will become an indispensable part of health care.
AI in healthcare has been praised for its great promise to diagnose and treat diseases, while its use in the medical field has been celebrated for the potential it brings to improve patient care. Doctors are using AI tools that can automatically read scans or images of the eye, detect abnormalities or cancerous tumors, and even identify skin conditions at a much faster rate than human doctors can do by themselves.
So, the use of AI in medicine and healthcare is growing. The medical field is changing radically to include more AI-powered tools, such as cognitive computing systems, which can help doctors make better decisions and quickly find answers to complex questions like “what are the symptoms of a heart attack?” or “what are the best treatments for breast cancer?”.
In recent years, there have been many new developments in the use of AI in medicine. As an example, researchers in 2016 have developed Deep Patient which is an unsupervised representation to predict the future of Patients from the Electronic Health Records.
In the figure is the conceptual framework used to derive the deep patient representation through unsupervised deep learning of a large EHR data warehouse. (A) Pre-processing stage to obtain raw patient representations from the EHRs. (B) The raw representations are modeled by the unsupervised deep architecture leading to a set of general and robust features. © The deep features are applied to the entire hospital database to derive patient representations that can be applied to a number of clinical tasks.
Following, is the diagram of the unsupervised deep feature learning pipeline to transform a raw dataset into the deep patient representation through multiple layers of neural networks. Each layer of the neural network is trained to produce a higher-level representation from the result of the previous layer.
AI may be able to reshape the field of healthcare — helping improve diagnostics and enabling a more personalized and precise approach to medicine — in almost limitless ways. Medical image quantification, automated genetic analysis, disease prediction, medical robotics, telemedicine, and virtual doctors are some of the main applications of AI in medicine. The coronavirus pandemic has accelerated the development and deployment of AI applications in the medical and clinical fields, as these technologies are at the core of the worldwide response to this health crisis.
You might also like to read Unlocking the Value of AI in Business Applications with ModelOps
Exploring the Different AI Applications in Healthcare
There is no doubt that AI is changing the landscape of healthcare. Some of the most common AI applications in healthcare include image recognition for diagnosis, patient monitoring, and drug discovery. However, the potential for AI in healthcare is far greater than these few examples.
One of the most exciting aspects of AI in healthcare is its ability to personalize treatments for each individual. With so much data available on patients, AI can analyze and learn from it to better understand how a person’s health history affects their current condition. This will allow doctors to provide more accurate diagnoses and treatments that are tailored.
Another field of application of AI is biomedical research, where the benefits can be multiple. AI can help researchers to find patterns in their data that they would have never seen before. It can also help them to better understand their data, which means that they will be able to come up with new hypotheses and ideas for experiments.
AI can undoubtedly help to shape the course of medical research.
To date, AI has gradually developed and introduced into nearly every field of medicine, from primary care to rare diseases, emergency medicine, biomedical research, and public health. Many administrative aspects related to healthcare administration (e.g. improving efficiency, quality control, reducing fraud) and policies are also expected to benefit from new AI-mediated tools.
- AI in clinical practice: there is tremendous potential for AI to be applied in the clinical setting, ranging from the automation of diagnostic processes to therapeutic decision making and clinical research. Data necessary for diagnosis and treatment come from a variety of sources, including clinical notes, laboratory tests, pharmacy data, medical imaging, and genomic information.
- AI in biomedical research: The use of AI-derived solutions in biomedical research appears to be more promising than in clinical applications, with recent advancements showing AI-derived solutions also proving useful in retrieving clinical knowledge. For example, established medical knowledge resources are already using ML algorithms to rank search results, including algorithms that continuously learn from user search behavior.
- Public health: a common definition of public health is ‘the science and art of preventing disease, prolonging life and promoting health through the organised efforts and informed choices of society, organisations, public and private, communities and individuals’ (Wanless, 2004). Experiments with related AI solutions are currently underway in several public health fields.
- AI in healthcare administration: the healthcare sector is characterized by an administrative workflow that involves a wide range of actors and institutions, including patients (e.g. billing management), health professionals, healthcare facilities and organizations (e.g. patient flow), imaging facilities, laboratories (e.g. consumable supply chain), pharmacies, payers, and regulators.
You might like to read also How ModelOps Helps You Execute Your AI Strategy
How to maintain regulatory compliance with AI in use?
When it comes to using artificial intelligence, many healthcare organizations are wondering how to ensure compliance with regulations such as HIPAA, GDPR, and AI Act to name a few. As AI becomes more common in the medical field, it is important that organizations understand the risks associated with its implementation and take steps to mitigate the risks.
Underestimating these aspects leads to many risks, which we can group into seven categories:
- Patient harm due to AI errors.
- Misuse of medical AI tools.
- Risk of bias in medical AI and perpetuation of inequities.
- Lack of transparency.
- Privacy and security issues.
- Gaps in AI accountability.
- Obstacles to implementation in real-world healthcare.
These risks not only harm patients and citizens but also reduce the trust of clinicians and society in AI algorithms. Therefore, risk assessment, classification and management must be an integral part of the AI development, assessment and deployment process.
The promise of artificial intelligence in medicine is to provide composite, panoramic views of individuals’ medical data; to improve decision making; to avoid errors such as misdiagnosis and unnecessary procedures; to help in the ordering and interpretation of appropriate tests; and to recommend treatment.
― Eric Topol, Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again
The Importance of the Models in Healthcare AI Research
AI in healthcare, as with other technological advances, also has specific benefits and risks and needs its own set of regulatory frameworks that address the socio-ethical implications of its use. While the implementation of AI in healthcare is very promising, being a rapidly developing field, it also raises concerns for patients, healthcare organizations and society. As we have already mentioned in the previous paragraphs, these concerns include issues of clinical safety, fair access, privacy and safety, appropriate use and users, as well as liability and regulation.
Therefore, stakeholders such as researchers, the general public and policymakers have all highlighted important bioethical issues which also include how to assess the risks and benefits of AI in healthcare, how to establish accountability in the biomedical AI sphere, and how to regulate its use in this particular high-risk context. Another important issue at the heart of the discussion is whether or not AI could increase inclusion and equity in the treatment of traditionally underrepresented communities, or whether it runs the risk of perpetuating and increasing pre-existing health disparities and inequalities.
We have already covered these extensively in our previous article “AI Act: A Risk-Based Policy Approach for Excellence and Trust in AI”, however, it is worth emphasizing the importance of implementing the Governance Model for AI that can mitigate risks and ensure the absence of biases for patients.
The specific characteristics of AI such as opacity, complexity, data dependence, and autonomous behaviour, can negatively affect a number of fundamental rights and the safety of users. Regulating the use of artificial intelligence models becomes a global concern to address these concerns. Therefore, we believe ModelOps is valuable in reducing business risk and ensuring that proposed models are free of errors and distortions. Companies preparing to develop artificial intelligence applications in the healthcare industry can benefit from ModelOps because it offers technical, business, and above all regulatory compliance controls and standardizes the model validation process thanks to a rules-based engine governing a large-scale artificial intelligence initiative.
In addition to government regulations, organizations utilizing artificial intelligence should meet the requirements set by the market. In the use of these technologies, consumers expect companies to adopt ethical and fair business practices, and therefore companies are called upon to comply. Furthermore, AI-based companies must ensure that they have governance policies and processes related to the models they use to address, at minimum, the following concerns:
- Efficacy: if models are not built or maintained properly, they can put the lives of individuals or entire companies at risk. To ensure models are robust and produce good results, they must be properly developed and validated periodically.
- Transparency: the decisions made by models need to be explainable and proven to be unbiased in order to protect individuals and groups from unfair treatment.
- Ethical use: artificial intelligence models, when developed improperly or with bias, can cause significant harm. For example, a healthcare application in the United States that uses AI to predict whether a patient will need multiple tests based on observing how the patient walks.
Despite the fact that the healthcare sector is highly regulated, no regulations cover the use of AI in healthcare settings. AI in health care has been the subject of proposed regulations by several countries and organizations, but no regulations have been adopted as of yet. So, the importance of regulating the use of AI models in the healthcare sector is a global concern. By using an Enterprise ModelOps Platform to Govern and Scale AI Initiatives, companies can:
- establishing clear policies regarding the standards that models must meet across all contexts, including business metrics, statistical metrics, internally generated compliance metrics, and external regulations.
- extensively documenting the purpose for each model, its metrics, how it was developed, and how it needs to be deployed.
- identifying all necessary approvals and approvers as the model moves from concept to development and into production.
- capturing all artifacts and metadata associated with the model, including code, training data, test cases, and results.
- logging all activities that happen with the model from the time of release to production through retirement, including deviations from KPIs, remediations such as code changes or retraining, and all approvals that took place at each step.
An example of capabilities and features of a well designed Enterprise ModelOps Platform.
According to most regulatory bodies, personnel who validate and test models must be separate from those who develop them. ModelOps platforms should include the continuously updated model inventory, which captures all the required documentation and metadata for every model, including any changes and approvals.
You might like to read also Modelops 2022: the state of practice
Conclusion & Summary of the Main Points
We have provided an overview of AI health-related applications and an analysis of the potential of AI to transform healthcare delivery. We have also seen what the risks are related to current and potential applications of AI in the healthcare sector. However, to address these socio-ethical, regulatory and technical issues, we explained how a ModelOps-based solution allows for the introduction of a number of options aimed at minimizing medical AI risks, strengthening governance at the Enterprise and therefore strengthening its responsible development.
- ModelOps for Dummies
- Modelop’s website
- Artificial intelligence in healthcare: Applications, risks, and ethical and societal impacts — EPRS (European Parliamentary Research Service) Scientific Foresight Unit (STOA)
- AI Act: A Risk-Based Policy Approach for Excellence and Trust in AI
- Deep Patient: An Unsupervised Representation to Predict the Future of Patients from the Electronic Health Records